apollo Routing Managing DomazniOS and

(1)

Managing DomazniOS and Domain Routing in an Internet

005694-AOO

apollo

(2)

Managing Domain/OS and Domain Routing in an Internet

Apollo Computer Inc.

330 Billerica Road Chelmsford, MA 01824

Order No. 005694-AOO

(3)

First Printing:

Latest Printing:

December 1985 July 1988

This do.cument was produced using the Interleaf Technical Publishing So.ftware (TPS) and the InterCAP Illustrato.r I Technical Illustrating System, a pro.duct o.f InterCAP Graphics Systems Co.rpo.ratio.n. Interleaf and TPS are trademarks o.f Interleaf, Inc.

Apo.llo. and Do.main are registered trademarks o.f Apo.llo. Co.mputer Inc.

ETHERNET is a registered trademark o.f Xero.x Co.rpo.ratio.n.

IBM is a registered trademark o.f Internatio.nal Business Machines Corpo.ratio.n.

MULTIBUS is a registered trademark o.f Intel Co.rpo.ratio.n.

UNIX and ACCUNET are registered trademarks o.f AT&T in the USA and o.ther co.untries.

VAX is a registered trademark o.f Digital Equipment Co.rpo.ratio.n.

3DGMR, Aegis, D3M, DGR, Do.main/Access, Do.main/Ada, Do.main/Bridge, Do.main/C, Do.main/Co.mCo.ntro.ller, Do.main/Co.mmo.nLISP, Do.main/CORE, Do.main/Debug, Do.main/DFL, Do.main/Dialogue, Domain/DQC, Do.main/IX, Do.main/Laser-26, Do.main/LISP, Do.main/PAK, Do.main/PCC, Do.main/PCI, Do.main/SNA. Do.main X.2S. DPSS, DPSS/Mail, DSEE, FPX, GMR, GPR, GSR, NLS, Netwo.rk Co.mputing Kernel, Netwo.rk Co.mputing System, Netwo.rk License Server, Open Dialo.gue, Open Netwo.rk To.o.lkit, Open System To.o.lkit, Perso.nal Superco.mputer, Personal Super Wo.rkstatio.n, Personal Workstation, Series 3000, Series 4000, Series 10000, and VCD-8 are trademarks of Apollo Computer Inc.

Apollo. Computer Inc. reserves the right to make changes in specifications and other information contained in this publication without prior no.tice, and the reader should in all cases co.nsult Apollo Computer Inc. to determine whether any such changes have been made.

THE TERMS AND CONDITIONS GOVERNING THE SALE OF APOLLO COMPUTER INC. HARDWARE PRODUCTS AND THE LICENSING OF APOLLO COMPUTER INC. SOFTWARE PROGRAMS CONSIST SOLELY OF THOSE SET FORTH IN THE WRITTEN CONTRACTS BETWEEN APOLLO COMPUTER INC. AND ITS CUSTOMERS. NO REPRESENTATION OR OTHER AFFIRMATION OF FACT CONTAINED IN THIS PUBLICATION, INCLUDING BUT NOT LIMITED TO STATEMENTS REGARDING CAPACITY, RESPONSE-TIME PERFORMANCE, SUITABILITY FOR USE OR PERFORMANCE OF PRODUCTS DESCRIBED HEREIN SHALL BE DEEMED TO BE A WARRANTY BY APOLLO COMPUTER INC. FOR ANY PURPOSE, OR GIVE RISE TO ANY LIABILITY BY APOLLO COMPUTER INC. WHATSOEVER.

IN NO EVENT SHALL APOLLO COMPUTER INC. BE LIABLE FOR ANY INCIDENTAL, INDIRECT, SPECIAL OR CONSEQUENTIAL DAMAGES WHATSOEVER (INCLUDING BUT NOT LIMITED TO LOST PROFITS) ARISING OUT OF OR RELATING TO THIS PUBLICATION OR THE INFORMATION CONTAINED IN IT, EVEN IF APOLLO COMPUTER INC. HAS BEEN ADVISED, KNEW OR SHOULD HAVE KNOWN OF THE POSSIBILITY OF SUCH DAMAGES.

THE SOFTWARE PROGRAMS DESCRIBED IN THIS DOCUMENT ARE CONFIDENTIAL INFORMATION AND PROPRIETARY PRODUCTS OF APOLLO COMPUTER INC. OR ITS LICENSORS.

(4)

Preface

Managing Domain/OS and Domain Routing in an Internet (005694-AOO) describes how to create, manage, and troubleshoot the SRI0 Domain® environment in an internet.

We've organized this manual as follows:

Chapter 1

Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Chapter 7 Chapter 8 Appendix A Appendix B

Appendix C Glossary

Provides an administrator's overview of the Domain environment in an internet and explains key internet management concepts.

Describes how to stage an internet installation.

Describes how to start the Domain routing process.

Describes how to create ns_helper, rgyd, and glbd in an internet Describes how to manage ns_helper in an internet.

Describes how to manage internet connectivity and performance.

Describes how to manage nodes in an internet.

Describes how to troubleshoot an internet.

Describes the network errors reported by our software.

Describes how to manage nodes that contain several network controllers.

Describes how to manage internets that contain SR9.n nodes.

Defines terms used in this book.

Preface iii

(5)

Related Manuals

The file /install/doc/apoll%s. v.latest software release number _manuals lists current titles and revisions for all available manuals.

For example, at SRI0.0 refer to /install/doc/apoll%s.v.l0.0_manuals to check that you are using the correct version of manuals. You may also want to use this file to check that you have ordered all of the manuals that you need.

(If you are using the Aegis environment, you can access the same information through the Help system by typing help manuals.)

Refer to the Domain Documentation Quick Reference (002685) and the Domain Documen- tation Master Index (011242) for a complete list of related documents. For more informa- tion on internets and Domain/OS, refer to the following documents:

• Planning Domain Networks and Internets (Order Number 009916-AOO)

• Domain Hardware Site Planning Specifications (Order Number 009858)

•

Making the Transition to SRlO Operating

System Releases (Order Number 011435)

•

^Managing^Aegis~System Software (Order Number 010852)

•

Managing SysV System Software (Order Number 010851)

•

Managing BSD System Software (Order Number 010853)

•

Managing the NCS Location Broker (Order Number 011895)

•

Making the Transition to SRIO TCPIIP (Order Number 011717)

•

Configuring and Managing TCPIIP (Order Number 008543-AOO)

•

Installing Software with Apollo's Release

and Installation Tools (Order Number 008860)

To manage nodes that run SR9.n software in your internet, refer to:

• Planning Domain Networks and Internets

• Installing Domain Software

• Managing the Domain Environment in an Internet

iv Preface

(Order Number 009916) (Order Number 008860) (Order Number 005694)

(6)

Problems, Questions, and Suggestions

We appreciate comments from the people who use our system. To make it easy for you to communicate with us, we provide the Apollo® Product Reporting (APR) system for comments related to hardware, software, and documentation. By using this formal channel, you make it easy for us to respond to your comments.

You can get more information about how to submit an APR by consulting the appropriate Command Reference manual for your environment (Aegis, BSD, or SysV). Refer to the mkapr (make apollo product report) shell command description. You can view the same description online by typing:

$ man mkapr (in the SysV environment)

% man mkapr (in the BSD environment)

$ help mkapr (in the Aegis environment)

Alternatively, you may use the Reader's Response Form at the back of this manual to submit comments about the manual.

Documentation Conventions

Unless otherwise noted in the text, this manual uses the following symbolic conventions.

literal values

user-supplied values

sample user input output

}

Bold words or characters in formats and command descriptions represent commands or keywords that you must use literally.

Pathnames are also in bold. Bold words in text indicate the first use of a new term.

Italic words or characters in formats and command descriptions represent values that you must supply.

In samples, information that the user enters appears in color.

Information that the system displays appears in this typeface.

Square brackets enclose optional items in formats and command descriptions.

Braces enclose a list from which you must choose an item in formats and command descriptions.

Preface v

(7)

< >

CTRLI

---88---

vi Preface

A vertical bar separates items in a list of choices.

Angle brackets enclose the name of a key on the keyboard.

The notation CTRLI followed by the name of a key indicates a control character sequence. Hold down <CTRL> while you press the key.

Horizontal ellipsis points indicate that you can repeat the preced- ing item one or more times.

Vertical ellipsis points mean that irrelevant parts of a figure or example have been omitted.

This symbol indicates the end of a chapter.

(8)

Chapter 1

1.1 1.2 1.3 1.3.1 1.3.2 1.3.3 1.3.4 1.4 1.4.1 1.5 1.5.1 1.6 1.6.1 1.6.2 1.6.3 1.6.3.1 1.6.3.2 1.7 1.8 1.9 1.10 1.10.1 1.10.2

The Domain Internet

Terminology and Conventions Used in the Text . . . . Supporting Documentation . . . . The Router . . . . Network Addresses . . . . The Domain Routing Algorithm . . . . The Maximum Hop Limitation . . . . The Router's Principal and Alternate Networks . . . . Domain Routing Service . . . . Levels of Internet Service . . . . Domain Services and Network Addresses . . . . Internet Configurations that Require Domain Services to Restart ... . Distributed Databases in Internets . . . . Domain Naming Service . . . . The NCS Location Broker . . . . The Domain Registry . . . . Selecting a Master Registry for the Internet . . . . Registry Summary . . . . Using Node Specifications in an Internet . . . . Using Pathnames in an Internet . . . . Using Shell Commands in an Internet . . . . Managing an Internet . . . . Changing Network Numbers . . . . Troubleshooting an Internet . . . .

1-2 1-3 1-4 1-4 1-6 1-9 1-9 1-10 1-11 1-12 1-13 1-15 1-16 1-17 1-17 1-18 1-19 1-20 1-22 1-22 1-24 1-24 1-25

Contents vii

(9)

Chapter 2 Staging an Internet Installation

2.1 Preparing the Site . . . .. 2-1 2.2 Planning for Shared Resources .-... 2-2 2.3 Coordinating the Internet Installation ... . . . .. 2-3 2.4 Shutting Down ns_helper Replicas . . . 2-4

Chapter 3 Starting the Routing Process

3.1 Verifying the Internet Configuration . . . .. 3-1 3.2 Assigning Network Numbers . . . 3-6 3.3 Enabling Routing Service. . . .. 3-8 3.4 Restoring the Command Search Rules . . . 3-9 3.5 Verifying the Routing Process . . . 3-10 3.6 Starting Routers from a Node Start-Up File ... 3-11

Chapter 4

4.1 4.2 4.3 4.4 4.4.1 4.4.2 4.5 4.6 4.7

Chapter 5 Preparing Domain System Resources

Preparing Database Servers . . . . Creating a Naming Database for the Internet ... . Merging Registries ... ~ ... . Creating a Global Location Broker for the Internet ... ,.

Creating the Global Location Broker From A Single Source ... . Creating the Global Location Broker From Disjoint Replicas ... . Updating SR10 Nodes for an Internet . . . . Following-Up Other Changes . . . . FiXing File Access Problems . . . .

Managing System Resources in an Internet

4-1 4-4 4-15 4-20 4-21 4-23 4-27 4-29 4-29

5.1 Using edns . . . 5-1 5.1.1 Adding Node Names to an os_helper Database. .. .. . .. .... . . .. 5-1 5.1.2 Adding Diskless Nodes. . . .. 5-3 5.1.3 Deleting Names from the Master Root Directory. . . .. 5-4 5.1.4 Replacing a Name in the os_helper, Database ... 5-4 5.1.5 Adding an os_helper Replica. . . .. . . .. 5-5 5.1.6 Maintaining and Repairing Replicas... . ... . .... .. .. . . .. .. 5-6 5.1.7 Stopping an ns_helper Replica . . . 5-7 5.1.8 Reinitializing an os_helper Database... . . . ... ... . ... .... .. 5-7

viii Contents

(10)

Chapter 6

6.1 6.1.1 6.1.2 6.1.3 6.1.4 6.2 6.2.1 6.2.2 6.2.3 6.3 6.4 6.4.1 6.4.2 6.4.3 6.4.4 6.5 6.6

Chapter 7 Managing the Internet Topology and Performance

Monitoring Domain Internet Topology . . . . Displaying Routing Tables . . . . Monitoring Current Topology . . . . Monitoring Topology Changes . . . . Monitoring Internet Availability . . . . Controlling Routers . . . . Displaying Routing Service . . . . Specifying Network Devices . . . . Setting and Changing Routing Service . . . . Assigning Network Numbers with the rtsvc Command ... . Collecting and Analyzing Routing Information . . . . Monitoring Internet Traffic . . . . Monitoring Device Usage . . . . Monitoring Device Operation . . . . Monitoring Performance at Timed Intervals . . . . Establishing Connections Without Routing . . . . Collecting Information about Individual Networks . . . .

Managing Nodes in an Internet

6-1 6-2 6-4 6-6 6-7 6-8 6-8 6-9 6-10 6-13 6-14 6-14 6-15 6-16 6-17 6-18 6-20

7.1 Changing the Network Number of All Nodes on a Network. . . .. 7-1 7.2 Moving a Node to Another Network . . . 7-7 7.3 Adding Another Network Controller to a Router. . . . .. 7-9

Chapter 8 Troubleshooting an Internet

8.1 Locating and Fixing Communication Failures . . . 8-1 8.1.1 Correcting Problems Caused by a Router . . . 8-2 8.1.2 Correcting Problems Caused by Incorrect Cached

Data or Naming Information . . . 8-4 8.2 Improving Performance . . . 8-6 8.3 Finding Problems from Remote Locations . . . 8-7

Contents ix

(11)

Appendix A Device Statistics

A.l ETH802.3_AT... . ... .... ... A-l A.2 ETH802.3_ VME ... A-2 A.3 IIC ... ... ... ... . ... ... A-3 A.4 RING. . . .. A-4

Appendix B Managing Network Controllers in a Gateway or Router

B.l B.2 B.2.1 B.2.2 B.3 B.3.1 B.3.2

Domain/OS and Network Devices ... . Network Configurations in Domain Nodes ... . Principal and Alternate Networks ... . Search Order for the Principal Network at Node Boot ... . Booting Networks in Service Mode ... . Finding a Boot Partner on an Alternate Network ... . Keeping the Configuration Tables Current ... .

Appendix C Creating An Internet of Mixed Operating System Releases

B-1 B-3 B-4 B-5 B-5 B-6 B-6

C.l Merging SR9.n and SR10 Registry Databases ... " C-l C.2 Updating SR9.n Systems for an Internet .' ... - C-2

Glossary Index

x ' Contents

(12)

Figures

1-1 1-2 1-3 1-4 1-5 1-6 1-7 1-8 1-9 1-10 1-11 1-12 1-13 1-14 2-1 3-1 4-1 4-2 4-3 4-4 4-5 6-1 6-2 6-3 6-4 6-5 6-6 6-7 8-1 8-2 8-3 8-4 8-5

Domain Environment in an Internet . . . 1-1 Reading Paths for System Administrators at SR10 ... 1-3 Assigning Domain Network Numbers. . . .. 1-5 Distributing Routing Information. . . .. 1-6 Router Broadcasting to Clients on Network A . . . .. 1-7 Multiple Routing Paths in a Domain Internet ... 1-8 Full Internet Routing Service - Open Ports. . . .. 1-10 Domain Services in New Internet. . . .. 1-12 Joining Separate Networks . . . 1-13 Partitioning a Network. . . .. 1-14 Adding Network to Existing Internet ... 1-14 Partitioning Network in Existing Internet. . . .. 1-14 Disjoint Databases before Merge . . . 1-15 An Internet with Point-to-Point Connection. . . .. 1-23 Diagram for Internet Installation . . . 2-3 Router Controller Schematics . . . 3-3 Creating an Internet Naming Database. . . .. 4-5 Adding New Registry to Existing Registry. . . . .. 4-15 Merging Registry Databases . . . .. 4-17 Creating glbdfFrom Single Source . . . 4-21 Merging Global Location Brokers. . . .. 4-23 Using the lcnet Command . . . 6'-3 Using Icnet -conn -hw . . . .. 6-4 Internet Described by lcnet -conn -hw ... 6-5 Notice from /sys/alarm/alarm_server -nets. . . .. 6-7 Effects of rtsvc -off ATR Side . . . 6-12 Effects of rtsvc -off IEEE 802.3 Side. . . .. 6-13 Using crp in an Internet . . . , 6-18 Fixing Internet Communications Problems. . . .. 8-3 Fixing Cached Data and Naming Problems ... .. 8-5 Local Failure Causing Internet Communication Problems ... 8-8 One and Two Hop Routing Paths. . . .. 8-10 Topologically Equivalent Internets . . . 8-11

Contents xi

(13)

Tables

Tasks

1-1 4-1 4-2 4-3 4-4 4-5 4-6 8-1 B-1

Inte~net Node Specificatidns ... . Nodes with Replicated Databases ... . Replica Comparison Table ... . List of Duplicate Node Names in Replica Databases ... . List of Changed Node Names ... . List of Changed Registry Account Names ... . Networks Where Unix Numbers Changed ... . Troubleshooting Problem List ... . Node Bus - Network Configurations ... .

Creating an Internet

1-21 4-2 4-8 4-10 4-13 4-19 4-19 8-12 B-4

1 Confirm the Router's Device Configuration ... 3-1 2 Confirm the Router's Network Configuration. . . .. 3-2 3 Shut Down Diskless Routing Nodes. . . .. 3-6 4 Change Command Search Rules ... 3-6 5 Assign Principal Network Number. . . .. 3-6 6 Delete the Routing Node's Hint File. . . .. 3-7 7 Reboot Routing Nodes ... 3-7 8 Verify the Router's Principal Network Number ... 3-7 9 Assign Alternate Network Number ... 3-8 10 Start the Routing Process. . . .. 3-8 11 Restore Command Search Rules ... 3-9 12 Verify Router Communication. . . .. 3-10 13 Verify Nonrouting Node Communication ... 3-10 14 Create Routing Start-Up Files. . . .. 3-12 15 Verify Routing Start-Up Files ... 3-12 16 Identify Nodes with Replicated Databases ... 4-2 17 Synchronize Node Clocks ... 4-3 18 Restart Server Processes. . . .. 4-4 19 Prepare Nodes for ns_helper Replicas ... '. . .. 4-6 20 Initialize ns_helper Replicas ... 4-7

xii Contents

(14)

Tasks (continued)

Creating an Internet (continued)

21 Compare ns_helper Replica Databases for Duplicate Names ... 4-8 22 Update One ns_helper Replica List . . . 4-10 23 Merge ns_helper Replica Databases. . . .. 4-11 24 Change Duplicate Node Names. . . .. 4-12 25 Verify Root Directory. . . .. 4-14 26 Create Names for Diskless Nodes. . . .. 4-14 27 Merge Registries. . . .. 4-17 28 Resolve Registry Collisions. . . .. 4-18 29 Update Master Registry Replica List. . . .. 4-19 30 Correct Replica Lists and Initialize New Replicas ... 4-21 31 Restart Replicas . . . 4-24 32 Correct Replica Lists . . . 4-24 33 Check Replica Clocks and Replica Availability ... 4-25 34 Merge glbd Databases . . . 4-26 35 Initialize New Replicas as Required. . . .. 4-26 36 Restart mbx_helper and spm . . . .. 4-27 37 Flush Local Naming Caches . . . 4-27 38 Update Local Registry . . . 4-28 39 Changes to Node Names . . . 4-29 40 Changes to Account Names. . . .. 4-29 41 Fix TCP/IP Network Addresses. . . . .. 4-29 Changing Network Numbers of All Nodes in a Network

1 Prepare Network to Return to 0 Address. . . .. 7-2 2 Stop Network Usage. . . .. 7-2 3 Shut Down ns_helper Replicas ... . . . .. 7-2 4 Isolate the Network . . . 7-3 5 Reconfigure the Physical Network. . . .. 7-3 6 Change Routing Startup Scripts . . . .. 7-4 7 Assign New Network Number . . . 7-4 8 Verify Router Communication. . . .. 7-4 9 Broadcast 0 Network Number. . . .. 7-5 10 Boot Nonrouting Nodes .... . . .. 7-5 11 Restore ns_helpers . . . .. 7-6 12 Change Registry . . . 7-6 13 Restore Global Location Broker . . . 7-6 14 Flush Local Naming Caches . . . 7-6·

15 Permit Network Usage . . . 7-7 16 Fix Any Remaining Communication Problems. . . .. 7-7

Contents xiii

(15)

Tasks (continued)

Moving A Node to Another Network

1 Prepare Node for New Environment. . . .. 7-7 2 Prepare Partner of Diskless Node. . . .. 7-8 3 Reconfigure the Physical Network. . . .. 7-8 4 Catalog the Node ... 7-8 Updating SR9.n Systems for an Internet

1 Restart mbx_helper and spm . . . .. C-2 2 Flush Local Naming Cache ... C-2 3 Update ACL Cache on SR9.n System ... C-3

xiv Contents

(16)

Chapter 1 The Domain Internet

Our network architecture supports the Domain environment in an internet. Apollo systems equipped with two or more network controllers and our standard software can transfer Domain packets between networks. Figure 1-1 illustrates Apollo nodes within an internet.

X.25 Network

Figure 1-1. Domain Environment in an Internet

In Figure 1-1, the physical networks that comprise the internet consist of an IEEE 802.3 network (commonly referred to as an ETHERNET* network) and two Apollo Token Ring

*IBM is a registered trademark of International Business Machines Corporation.

VAX is a registered trademark of Digital Equipment Corporation.

ETHERNET is a registered trademark of Xerox Corporation.

The Domain Internet 1-1

(17)

networks. Apollo workstations use Domain protocols to communicate with each other, and a variety of other protocols to communicate with other vendors' systems.

The Domain environment in an internet is a single, transparent system. Users need not be concerned with object locations or network addresses. As in a single network, Apollo workstations anywhere in the internet provide each other with Domain services such as:

• A distributed file system and network-wide naming directory

• Transparent resource sharing, e.g., paging over network links

• Remote login and remote process creation

• Communications services to other systems and networks

One Domain service, diskless booting, is not provided over an internet. Diskless nodes and their booting partners must be on the same physical network.

1.1 Terminology and Conventions Used in the Text

Although the term internet is generic, this book uses a more explicit definition. Domain internet describes a collection of networks that use Domain network addresses and communication protocols. The term routing refers to methods of finding and using available pathways from one network to another. Here, the term describes Domain routing service exclusively.

The terms server or daemon refer to software processes that perform specific functions on behalf of clients. Naming service describes an internet facility that mediates host names and their network addresses. Here, naming service applies to /sys/ns/ns_helper, a server that manages Apollo Domain node names and their Domain network addresses. Note that in this book, the term Domain name never refers TCP/IP domain names. This manual's glossary gives additional information about many of the terms used in the text.

We refer collectively to the system administrator's manuals for each Domain/OS environment as system administrator's manuals or guides. The Preface contains a list of titles.

You can perform the tasks described in this book in any Domain/OS environment. We use the dollar sign ($) to indicate the shell prompt, however $ does not imply a particular shell type. With few exceptions, the commands shown after the shell prompt work in all

Domain/OS shells. We specify commands for particular Domain/OS environments when necessary.

We use the convention 'node_data in examples that refer to the directory containing a node's data files. In Unix environments, you must precede the ' (tick) character with a \

(backslash) to refer to \ 'node_data.

1-2 The Domain Internet

(18)

1.2 Supporting Documentation

Use Planning Domain Networks and Internets to plan an internet configuration and to select the hardware to make network interconnections. The book describes physical configurations and hardware used by our systems for internet connections.

The information in Planning Domain Networks and Internets applies to any family of internet protocols supported by our standard software. Currently, these protocols are the Domain internet routing protocol and TCP/IP. Managing Domain/OS and Domain Routing in an Internet describes how to extend the SR10 Domain/OS over several networks.

Figure 1-2 shows some reading paths for system administrators who manage internets run- ning SR10 Domain/OS.

Planning Domain Networks and Internets Domain Hardware Site Planning Specifications

Making the Transition to SR10 Operating System Releases Installing Software with Apollo's Release and Installation Tools

Managing Domain/OS and Domain Routing in an Internet

Managing SysV System Software Managing BSD System Software Managing Aegis System Software Managing the NCS Location Broker

Making the Transition to SRlO TCPIIP

Configuring and Managing TCP /IP

Figure 1-2. Reading Paths for System Administrators at SRlO

This chapter will help you to become familiar with the Domain operating environment in an internet. It contains explanations of important concepts that you must understand in order to create and manage the Domain environment in an internet. Please review the material in this chapter before you create an internet.

The Domain Internet 1-3

(19)

1.3 The Router

Routing is the process of sending packets from one network to another. We refer to the nodes that send Domain packets to other networks as routers. Routers connect to several networks simultaneously. Our terminology emphasizes the router's primary function, for- matting packet headers and forwarding packets on behalf of other Domain nodes which are clients of the Domain routers.

The term gateway also describes processors that send packets to other networks. However, gateways usually perform protocol translation to assist communication between unlike systems. At most, Domain routers perform media conversion functions. Since Domain routers assist communication between like systems, protocol translation is not necessary.

A Domain node's system bus can contain as many as four network controllers although some configurations permit no more than two network connections in a bus. A router's system bus contains at least two network controllers. Appendix B describes how to manage nodes that contain several network controllers.

As part of a Domain node, our network controllers receive and transmit signal streams on the networks to which they connect. The controllers create and recognize meaningful data structures within the signal stream; they ensure error free transmission and respond appro- priately to errors that do occur. Controllers report statistics and errors about the networks to which they are attached. Appendix A describes network device statistics. You can moni- tor, troubleshoot and service the controllers installed in routers using the same methods that apply to nodes that contain one controller.

To create a Domain internet, configure nodes with the appropriate network hardware and start Domain routing service. Once you start the routing process, it maintains itself automatically.

1.3.1 Network Addresses

A Domain router needs our standard software revision 9.5 or later to send packets to other networks. Every routing scheme uses a network address to identify networks within the internet. The content and format of the address depends on the particular internet protocol in use.

You must assign an 8-digit hexadecimal network number to each network that is part of the Domain internet. This number is the network's unique address and distinguishes it from all other Domain networks in the internet.

NOTE: You can ass,ign your own numbers or obtain them from us. We guarantee that our network numbers are unique. Chapter 2 contains information about obtaining network numbers from us.

(20)

The Domain internet is the collection of networks you can access with Domain network numbers. Most of the networks that comprise a Domain internet contain Domain nodes.

Some of the networks may be connecting links that do not contain any Domain nodes other than routers. Nevertheless, you must assign an address to each network whether it contains Domain nodes or is simply a connecting link.

You can run multiple internet protocols over the same physical internet. For example, TCP/IP can run in parallel with Domain routing on the same nodes and/or in the same networks. However, TCP/IP and Domain internets use different addressing schemes, are implemented in different ways, use different management utilities, and provide different facilities for network users.

If your internet is configured so that multiple protocols run in parallel, assign Domain network numbers to all the networks through which Domain packets pass. It isn't necessary to assign Domain network numbers to networks that do not contain Domain nodes, or do not serve as connecting links.

For example, Figure 1-3 shows an internet that contains Domain nodes on networks A and B. Domain routers 1 and 2 connect to network C which serves as a point-to-point link between networks A and B. Domain communication occurs on networks A, Band C.

There is no Domain communication to network D. In this internet, you assign Domain network numbers to networks A, Band C; do not assign Domain network numbers to network D.

VAX

Figure 1-3. Assigning Domain Network Numbers

However, systems anywhere in this internet can use TCP/IP to communicate. For example, Domain nodes in networks A and B communicate with the VAX system in network D by using TCP/IP. As a consequence, all networks in this internet have TCP/IP network addresses. Nodes 1 and 2 function simultaneously as Domain routers and TCP/IP gateways.

The Domain Internet 1-5

(21)

1.3.2 The Domain Routing Algorithm

Domain routers and their clients use a distributed routing algorithm. Each router maintains a routing table containing the addresses of the networks to which it connects and the number of hops to each network in the internet. (A hop is a measure of distance between networks. Each time a packet passes through a router, it makes one hop.)

Periodically, routers broadcast the information in their routing tables to their local networks. Routers that connect to the same network learn about routes to new networks when they receive these broadcasts. Figure 1-4 illustrates an internet that contains two Apollo Token Rings (A and C) and one IEEE 802.3 network (B).

Figure 1-4. Distributing Routing Information

Router 1 connects to networks A and B, router 2 connects to networks Band C. Both routers broadcast their routing information on network B. Thus, router 1 learns about a path to network C and router 2 learns about a path to network A.

Whenever they receive information about a new network, routers revise their routing tables and include the new information. Eventually, every router has paths to all the networks in the internet. The Icnet command (Iist_connected_networks) displays information in routing tables. (Subsection 6.1. 3, "Monitoring Topology Changes," contains more information about routing tables and how they are revised.)

Client nodes learn their own network addresses and the number of hops to other networks from routing broadcasts to their local networks. Clients also keep routing tables. Figure 1-5 shows the information router 1 broadcasts to network A, and the information the client node puts in its routing table. (Note that Figure 1-5 doesn't show similar information that router 1 broadcasts to network B.)

Domain packet headers contain three address fields. The To field contains the address of the packet's target, the From field contains the address of the source, i.e., the return address, and the Via field specifies the first hop on the route to the destination address.

Domain routing is adaptive, that is, packets do not travel ^toother networks via scheduled paths. Instead, packets travel via the shortest route measured in hops.

1-6 The Domain Internet

(22)

B

--~=-2

^C

My address is Network A Network B is 1 hop via 1 Network C is 2 hops via 1

A

Figure 1-5. Router Broadcasting to Clients on Network A

When a node is about to transmit a packet, it looks up the target network in its routing table. If the network is local, the source node sends the packet via the target node on the local network, i.e., the Via field and the To field contain the same address. If the target is on another network, the source sends the packet via the router that is the first hop to the destination. The packet's To field and Via field contain different addresses. In the internet shown in Figure 1-5, nodes in network A send packets to network C via router 1.

Once the node assembles the packet header information, it selects a network controller to transmit the packet, and the controller puts the packet on the network. Routers can select one of several network controllers; clients must use their single controller.

When the packet reaches the first hop, the adaptive routing scenario is re-enacted. The router examines the packet's To address field, looks in its routing table for the first hop to the destination, puts a new address in the Via field, selects a network device, and the packet is transmitted.

In Figure 1-6, the nodes that provide routing service are labeled 1, 2 and 3. Routers 1 and 2 connect to the IEEE 802.3 network (B) and to an Apollo Token Ring network (A and C). Router 3 connects to two ATR networks (C and D) and to the IEEE 802.3 network (B).

This internet contains redundant routing paths. For example, packets traveling between network C and network A can go through router 2 or router 3. In either case, it is two hops from network C to network A.

(23)

A

Figure 1-6. Multiple Routing Paths in a Domain Internet

Router 1 is the first hop for packets sent from network A to network C. Its routing table contains information about one of the available paths to network C. The routing table can look like either of the following lists:

Network A is local Network B is local Network C is 1 hop via 2 Network D is 1 hop via 3

Network A is local Network B is local

Network C is 1 hop via 3 Network D is 1 hop via 3

When there are multiple paths of the same distance to some network, routers send packets by the first path they hear about. Routers continue to use that path unless they receive information about a shorter path. Routers always use the shortest path no matter how busy that path might be.

When an internet contains several routing paths of the same distance, packets can be sent and returned via different routers. For example in Figure 1-6, router 1 can send packets from network A to network C via router 2, and the target node in network C can return the packet via router 3. The round-trip route depends on the information that is in the routing tables at the source, the router, and the target.

The important information to note however, is that the number of forward and return hops is equal. The distance from A to C and the distance from C to A is the same: two hops.

You can often pinpoint problems in the internet if you see that two nodes have different information about the distance between them. Section 8.3, "Finding Problems from Re- mote Locations," contains more information about such problems.

In Domain internets, nodes acknowledge receipt of a packet by sending back the information requested. If the sending node gets no reply to a message within an expected time frame, its operating system sends the message again. After several retries, the sender reports a failure. When this happens, the user sees a standard error message from the operating system, for example, "Remote node failed to respond" or "Object not found."

(24)

1.3.3 The Maximum Hop Limitation

The total number of hops between any two nodes depends on the internet's configuration.

The maximum number of hops allowed in a Domain internet is sixteen, provided that every network in the internet meets minimum network throughput performance guidelines.

The minimum transmission rate over any single link must be at least one megabit per sec- ond with a maximum end-to-end delay of 100 milliseconds.

Our guidelines ensure that a packet will traverse the internet within the time specified by Domain protocols. As your internet configuration approaches the limits of our guidelines, it can perform in an unsatisfactory manner: Some configurations that exceed our limits can work, but their performance and reliability will be reduced.

To improve internet performance in any configuration, you c~n reconfigure segments within networks, the number of hops between networks, or remove or replace equipment that slows data transmission. Refer to Planning Domain Networks and Internets for descriptions of internet configurations and equipment to use to make your internet perform adequately.

1.3.4 The Router's Principal and Alternate Networks

Apollo workstations initialize a network controller as part of their boot sequence. The network controller initialized at node boot is the node's principal network. By default, network services (for example, diskless boot, registry, and naming services) are supplied from servers on a node's principal network. A diskless node boots automatically on its principal network.

Routers contain several network controllers. Like all Domain nodes, the controller initialized at system boot is the router's principal network. All other controllers in a router sup- port alternate networks. You initialize alternate network controllers when you start routing service.

Domain/OS displays a device name for all network controllers installed in a node. For example, shell commands display the device name ring for Apollo Token Ring controllers.

If a router contains devices of the same type, the system differentiates them as unit 0 ... unit n, for example ring 1.

The principal or alternate status of a network is of little concern when a router is operating. You can specify either the principal or alternate network in shell commands. If the system crashes, Domain routers retain their principal network number. Typically, you place network numbers in the router's startup scripts because a router does not retain alternate network numbers after a crash.

The partner of a diskless router must be on the router's principal network. If you change a diskless router's principal network, you must edit /sys/net/diskless_Iist on a partner in the new principal network. Whenever possible, decide on a node's hardware configuration in

(25)

advance, then plan to stay with that configuration for some time. See Appendix B for more information about managing a router's principal and alternate networks.

1.4 Domain Routing Service

Domain nodes offer services to each other directly through a request-response protocol.

Nodes act as clients when they requests services directly from another; nodes act as servers when responding to requests from other Domain nodes. Under this model, all Domain nodes can request and provide the following services to each other immediately after booting: remote process creation, bootstrap, paging and file system service, node and network status reporting.

Domain nodes can provide an additional set of services, if you configure them to do so.

Typically, one node offers a specific service to other nodes who are clients of that service.

Some examples of these services are Domain naming, registry and routing services.

Domain routing service is part of our standard software. You can enable routing at nodes configured with two or more network controllers. Routers pass requests for Domain service from Domain nodes on one network to Domain nodes on other networks. Because the node provides this one service, we refer to it as a router. At the same time that a node acts as a router, it can also request and provide any other types of services offered on Domain nodes.

The routing process uses a software interface called a port to gain access to the networks.

As shown in the schematic in Figure 1-7, routers have one port for each connected network. A port transfers packets between the operating system and a network controller. The controller performs the interface functions necessary to get data physically on and off the connected network. You assign network numbers to the node's network ports.

ROUTER A

OPERATING SYSTEM

NETWORK 1230 NETWORK 1231

OPERATING SYSTEM

ROUTER B

NETWORK 1232

Figure 1-7. Full Internet Routing Service - Open Ports

(26)

Domain/OS uses the rtsvc command (routing_service) to control routing service. Use rtsvc to enable or disable Domain service to networks in the internet. Chapter 3 describes how to enable routing service when you create an internet. Chapters 6 and 8 describe how to control routers in existing internets.

You can control Domain internet routing paths by using rtsvc command options. You can isolate networks in the internet for software testing, network troubleshooting, or equipment installation. If your internet contains redundant paths to some networks, you can direct packets over different paths by stopping Domain routing service at a given port. Note that whenever you manipulate routing paths, the adaptive routing scheme is still in effect. The packet always travels by the shortest available route.

1.4.1 Levels of Internet Service

You can allow the following types of services at a port:

• Full Domain routing service (rtsvc -route). The operating system accepts internet packets on behalf of clients. Domain packets can pass to and from other ports in the node. A node becomes a router when you enable routing services at two or more of its ports. Rtsvc -route is the option usually in effect on active routers.

• Domain service on behalf of this node only (rtsvc -noroute). The operating system does not route on behalf of clients. However, the network ports are open for Domain service on the local network and for non-Domain packets that can use the port. Rtsvc -noroute can be used for some internet management tasks described later in this manual.

• No network services (rtsvc -off). The operating system does not make or accept requests for any Domain service on behalf of itself or others. Typically, you use rtsvc -off during internet installations and troubleshooting.

When you issue the rtsvc -noroute command, a router cannot provide routing service for other nodes. However, the node can request Domain service for itself. It can provide Domain services other than routing to nodes on its local network. For example, you can use the node to read files located on remote nodes. The node can respond to requests from other Domain nodes; for example the node can provide topology information.

The rtsvc -noroute command does not effect the node's ability to act as a server on its local networks; for example the node can offer registry or naming service. The -noroute option simply stops Domain routing, it does not affect any other service.

Routers are always clients of other routers on their local networks no matter what level of routing service they provide. When rtsvc -noroute is in effect on a particular router, the node can communicate with a node in another network if a routing path to that network is available. In other words, the node can become the client of another router although it cannot route on behalf of other nodes.

(27)

In summary, Domain nodes can request and provide many kinds of services. All nodes can provide some services automatically by virtue of their connection to the network. You implement specific services, including routing service, on selected nodes. A node can provide routing service and other kinds of Domain service simultaneously.

You control Domain routing service with the rtsvc command; you control the access that other Domain services have to a node's network ports with the netsvc command (net- work_service). Section 6.2, "Controlling Routers," contains important information about the effects of these commands on routers.

1.5 Domain Services and Network Addresses

Processes that provide Domain services do not have direct access to routing broadcasts.

Processes learn the node's principal network address as they initialize. All Domain networks use 0 as their network address until you assign a unique network address. When you start routing, processes that are currently executing continue to use 0 as their network address. This address is sufficient for service on the local network. However, you must restart processes that provide Domain service in order to use them in an internet.

For example, Figure 1-8 shows the messages the router IInew_york broadcasts to its local networks (A and B) in a new internet. Domain processes have not been restarted. At this point, networks are unstable.

ffrome

Figure 1-8. Domain Services in New Internet

For example, a user at Illondon can create a remote process on Ilparis in network A but cannot do the same on IIrome in network B. The serveryrocess_manager, Isys/spm/spm,

(28)

manages remote process creation. The spm processes executing on //Iondon, IIparis and / /rome continue to use a 0 network address even though routers are broadcasting unique addresses. Should IIlondon attempt communication with //rome, the packet never leaves network A and communication fails.

Moreover if //london reboots and /lparis does not, some Domain communications between these nodes can fail even though the nodes are on the same network. When / /Iondon attempts to create a remote process on /lparis, the packet arrives at IIparis, but inter- process communication can fail. Alternatively, //paris might receive and service a request from //london, but the packet might not return.

In summary, when you create an internet there is an unstable period that temporarily effects service. When Domain processes restart with their unique network addresses, users can create processes anywhere in the internet.

1.5.1 Internet Configurations that Require Domain Services to Restart

You must restart Domain services in any network that receives a new network number.

Figure 1-9 illustrates a Domain internet that is created by joining two networks that use 0 as their network address. Domain services in networks 1230 and 1231 must be restarted with their new network numbers.

Figure 1-9. Joining Separate Networks

Figure 1-10 illustrates a Domain internet created by partitioning a single network into smaller networks. The single network uses 0 as a network address. After the partition, Domain services in networks 1230 and 1231 must be restarted with their new network numbers.

(29)

!!cat

II

Figure 1-10. Partitioning Network

Figure 1-11 shows an internet containing networks 1230 and 1231. A new network with address 0 is added to the existing internet. Network 0 becomes network 1232. Domain services are restarted in network 1232; however it isn't necessary to restart services in networks 1230 and 1231.

!!sheep

...

. ^~

o

^~

..

Q

^#::

IIgoat

Figure 1-11. Adding Network to Existing Internet

Figure 1-12 shows a Domain internet created by partitioning a network within an existing internet. In this configuration, network 1231 is partitioned. After the partition, some nodes that were part of network 1231 are now part of a new network, 1232. Domain services restart only in the networks that receive a new network number; so in this example services restart on nodes in network 1232. Chapter 4 contains procedures for restarting Domain services in a variety of internet scenarios.

!!cat

Figure 1-12. Partitioning Network in Existing Internet

(30)

1.6 Distributed Databases in Internets

Some Domain services have databases that clients must use in order to execute tasks. For example, nodes use the Registry Server database to validate login. You can distribute replicas, multiple occurrences of a database, throughout a network to ensure continuity of service. Replicas automatically attempt to maintain consistency among their databases.

When you join two networks, databases that were operating independently are disjoint. You merge them in order to make them consistent. For example, if you cable two Apollo Token Ring networks together to make one larger network, the ns_helper, registry, and glbd in the new network are disjoint. Once you merge these databases, the replicas can continue to maintain consistency automatically.

When you join separate networks into an internet, add a network to an existing internet, or join several internets into a single larger internet, distributed databases are disjoint and must be merged. In addition, databases that were operating independently receive a new network address. Replica lists must be updated to include the new network address; then, replicas can maintain consistency automatically.

Figure 1-13 shows a new internet created by joining network X and network Y. The registry databases have not been merged. A. Random User can log in successfully in network X because the registry database in network X contains this account. A. Random User cannot log in successfully in network Y; the registry database does not contain the user's account.

When the registries are merged and the replica lists updated, A. Random User can log in from anywhere in the internet.

mozart.%.%

bach.%.%

handel.%.%

IIregistry_y

a random user. %. % copernicus. %. % gallileo.%.%

Figure 1-13. Disjoint Databases before Merge

(31)

If you create an internet by partitioning a single large network into several smaller networks, distributed databases are not disjoint. In such an internet, all replicas are instances of the database in the original network. The following sections describe the services that use distributed databases and provide additional information that applies to their use in an internet.

1.6.1 Domain Naming Service

The Isys/ns/ns_helper (naminLserver_helper) maintains the network root directory (II), a database containing the entry directory (I) name, node_ID and network address of all Domain nodes. Clients' root directories are a subset or cache of the network root directory. When a node cannot find an object using information in its cache, it queries the ns_helper database. The node updates its cache with new information it receives from ns_helper.

In addition to names and network addresses, each ns_helper database contains a replica list of all ns_helper processes. Whenever an ns_helper gets new information, it propagates that information to the replicas on the list. By this means, ns_helper maintains consistency among replicas' databases.

You can choose to implement ns_helper in a single network, but you must implement ns_helper in a Domain internet. In an internet, ns_helper's database contains information about Domain nodes in all networks. However, ns_helper serves clients on its local network only; it cannot serve nodes in other networks. Therefore, at least one ns_helper process must run on each network in the internet.

When you implement ns_helper in an internet, you merge the databases of existing replicas to ensure that each replica has a complete internet root directory and a complete list of all replicas in the internet. Once replica lists are complete, you can add new information to any ns_helper and it propagates that information to replicas throughout the internet.

If you have not used ns_helper in a single network, we recommend that you read about this service in your system administrator's guide and keep your guide available as you perform the tasks in Chapter 4. However, we do provide step-by-step procedures in this book so that you can implement naming service in an internet, even if you are not very familiar with ns_helper.

In summary, when you move ns_helper into an internet, you recreate existing replicas in networks that receive new network numbers. You merge existing replicas databases to ensure that replica lists and node names are complete and inclusive of the entire internet.

Each network in the internet must have at least one ns_helper replica. If you haven't used ns_helper in your network(s), our procedures tell you when to start replicas as you create the internet.

(32)

1.6.2 The NCS Location Broker

The Network Computing System ™ (NCS) is a set of tools that supports both Domain and non-Domain nodes' abilities to distribute computing on a network. One NCS facility, the Location Broker maintains information about the identities and locations of NCS resources in a network. Applications that require NCS resources query the Location Broker as their needs arise. The Global Location Broker (GLB) database contains information about all NCS resources in a internet.

The GLB executes as the daemon letc/ncs/glbd; the glbd daemon can be replicated throughout a network or internet. The Data Replication Manager (DRM), a server that executes with every glbd, manages glbd replicas. You can use the DRM to modify a GLB replica list. In addition, the DRM propagates new information in the database to replicas on its list of glbd replicas.

The Local Location Broker (LLB) executes as the daemon letc/ncslllbd on every node that contains NCS servers. The Ilbd maintains a database of the NCS resources residing on the host and it can forward client requests to objects on the host. An Ilbd executes on every node that contains a GLB because each GLB is itself an NCS resource.

Note that the GLB provides information to clients on its local network. It cannot provide information to clients on other networks. Each network in an internet must have at least one glbd. When you configure your internet, you must merge the glbd databases in networks that operated separately. In addition, glbd replica lists must include the correct names and addresses for all glbds.

When you create an internet, you must restart glbd in networks that receive new network numbers. Procedures in Chapter 4 tell you when and how to restart glbd. In addition, these procedures describe how to merge glbd databases and DRM replica lists, and when and how to restart Ilbd on NCS hosts.

Managing the NCS Location Broker fully describes the Network Computing System's administrative facilities. If you are not familiar with the NCS administrator's tools, we recommend that you have that manual available as you perform the tasks in Chapter 4. It contains detailed explanations of the administrator's commands and their operation. How- ever, in this book we do provide step-by-step procedures so that you can complete the required tasks, even if you are not familiar with NCS administrator's tools.

1.6.3 The Domain Registry

A Domain registry is a distributed database that identifies legitimate users of a secure Domain network. The operating system uses the registry to validate a user's attempt to log in. Each internet has one master Registry Server (letc/rgyd) that manages a database of names and accounts (lsys/registry/rgy_data).

The Domain Internet 1-17·

(33)

A network may contain any number of replicas of the Registry Server. The replicas are slaves of the master registry. Slave registries can validate login. If changes to the registry database are necessary however, only the master registry can be edited directly. The master propagates changes in its database to slave registries. Refer to any of the system administrators' guides for complete information about the operation of the Domain registry.

If you create an internet by partitioning one network into several networks, the existing master registry can continue to serve as the master registry in the internet. However when you create an internet by joining separate networks, a master registry exists in each network. When you add a network to an existing internet, both the internet and the new network have a master. Likewise, if you join separate Domain internets into a single internet, each internet has a master registry.

One and only one Registry Server can become the master in the new internet configuration. As we explained in Section 1.6, "Distributed Databases in Internets," you must merge the registry databases so that users can log in from anywhere in the internet. Only master registries can be modified directly, so you merge the master registries from the networks that form the internet.

The rgy_merge command merges master registries. You execute rgy_merge at a target.

Other master registries are sources; their databases are merged into the target. The target becomes the master registry in the internet, the sources become slaves.

For example, suppose you join network A and network B and you want the master registry for the internet located in network A. Execute rgy_merge on the master in network A to make it the master for the internet. The master on network B becomes a slave in the new internet. Any other slaves that exist in either network A or B become slaves of the new internet master registry. Chapter 4 describes how to merge registries when you form an internet.

1. 6. 3.1 Selecting a Master Registry for the Internet

The primary requirement for the internet master registry is continuous availability. Both the node and the network that contains the master registry should be available at all times. In addition, nodes that run the Registry Server must have disks and at least 3Mb of physical memory. If possible, select a master with a large database as the target. The registry merge operation completes faster when you merge small sources into a larger target.

The registry allows you to set security policy. For example, you can regulate the lifespan of passwords. If the registries that you merge have differing policies, rgy_merge preserves the policy in effect at the target of the merge. This is true even when the policy at the target is less restrictive than the policy at the source.

Thus, the policy in effect for the internet depends on which master you've chosen as the target. That policy might be acceptable, or you might want to use one of the policies in effect at a source, or set an entirely different policy for the internet.

(34)

From the administrator's point of view, merging registries with different policies and changing the policy on an existing registry are the same. Simply use edrgy -prop to set the registry policy. The system administrator's guides contain information about changing registry policies.

Users whose passwords do not conform to the policy in effect after the merge are able to log in. The system checks policy only at "change-of-password" times which are

• Password expiration date

• Password lifespan expiration

• Explicit changes made with chpass or edrgy

At such times the system notifies users to change passwords that do not conform to the internet registry policy. If this scenario is acceptable, you need do nothing to the registry policy. Alternatively, use edryg -prop to set a password expiration date that takes effect, for example, in 24 hours. Then passwords must change to be in compliance with the internet registry policy.

The policy that applies to an organization at the target is preserved after the merge. For example, both target and source have the default organization, sys_org. The policy that applies to sys_org at the target becomes the policy for any sys_org merged into the target.

When an organization exists in the source but not in the target (it is a new organization at the target), its policy remains in effect at the target.

1.6.3.2 Registry Summary

In summary, when you move the registry into an internet, select a node and network that is continuously available as the location for the master registry. Merge existing masters into the database that you've selected as the internet master. After the merge completes, the former masters and their slaves become slaves of the master in the internet. Security policy in effect at the internet master applies to the entire internet after the merge.

The Domain Registry is an NCS resource. Because Domain nodes anywhere in the internet can find a registry through the Location Broker, it isn't absolutely necessary that each network contain a registry slave. We recommend that you place a slave in any network that has a single routing path to the rest of the internet. In this way, registry service is continuously available and is not dependent on the availability of routing service. When networks have several routing paths to the rest of the internet, examine your internet configuration to determine the best locations for each slave.

The Domain Internet 1-19